Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

MiniMax loses key engineering lead after M3 launch: Skyler Miao departs

MiniMax's Agent Engineering department lead, Ada (Skyler Miao), has left the company, according to Blue Whale News, which cited his Feishu status now showing departure; his next destination is not yet known. Miao described himself on X as MiniMax's Head of Engineering, responsible for M2.x, Agent, Audio, and Hailuo AI, joining in 2023 after stints at Baidu, Beike, and ByteDance.

Published

MiniMax has seen a significant personnel change shortly after the release of its M3 model. According to a Blue Whale News report citing public information, Ada (Skyler Miao, 缪宇航), head of MiniMax's Agent Engineering department, has left the company — his Feishu status now shows as departed, and his next destination has not been announced. His X profile verification has not yet been updated either.

On X, Miao introduces himself as MiniMax's Head of Engineering, responsible for M2.x, Agent, Audio, and Hailuo AI, with his latest post being an AI-related retweet from August 3.

His public interactions show deep involvement in technical communication around the M3 launch: he teased MiniMax's Sparse Attention work, answered developers' questions about it, and said the implementation code would be open-sourced; when M3 shipped, he wrote that the team had gone through many sleepless nights.

Before joining MiniMax in 2023, Miao worked at Baidu, Beike, and ByteDance. At MiniMax he contributed to models, Agent products, and Hailuo AI, and long served as the company's technical explainer and outward-facing communicator with the developer community.

The departure lands at a delicate moment: M3 has just shipped, and the company is still pushing the open-sourcing of Sparse Attention and its ecosystem building. How the loss of this core engineering lead affects M3's follow-up iterations, the Agent product line, and Hailuo AI's development pace is now the question to watch.

The report's headline — “after M3's release, MiniMax lost a key general” — underlines how much weight this role carried. The next things to watch are where Miao lands next and whether MiniMax offers more details on handovers, the Sparse Attention open-source plan, and the Agent engineering team's roadmap.

Why it matters

Losing the engineering lead responsible for models, Agent, and Hailuo AI right after M3's launch creates uncertainty around MiniMax's product cadence and developer outreach, and his next move will shape how the market reads the company's trajectory.

MiniMaxM3Talent
Back to realtime news

Nearby Updates

All

08/21, 15:17

Generalist AI releases GEN-1.5 robot foundation model that learns new skills from 3-second demos

Robot startup Generalist AI has released GEN-1.5, a foundation model that learns new manipulation skills from just 3-12 second demonstrations, with zero gradient updates or fine-tuning. The model can also chain multiple demos into continuous tasks and transfer virtual demonstrations to the real world, prompting comparisons to a GPT-3 moment for robotics.

08/21, 16:03

Ping An reports AI agents drove 57.3B yuan in sales in H1 2026 as daily token usage hits 120B

Ping An has released its first-half 2026 results, reporting that AI agents helped generate more than 57.3 billion yuan in sales. The company's AI systems now consume 120 billion tokens per day, a scale that shows large models deeply embedded in its insurance and financial operations.

08/21, 13:59

SenseTime open-sources 8B multimodal model with native 4K image output

SenseTime has open-sourced an 8B-parameter multimodal model with native 4K image output. The release brings ultra-high-resolution image generation into the open-source community and gives developers a new multimodal foundation model option.

08/21, 13:33

DP Technology moves the full research workflow onto a desktop: scientists ask questions, AI runs experiments

DP Technology has introduced a desktop product that moves the full scientific research workflow onto a single desktop, where scientists ask questions and AI runs the experiments. The company says the goal is to give scientists' time back to scientific creation.